GeezerVison

I've refreshed this PR with a newer upstream.

• Eliminates an unused variable.
• Switches to a saner API of finish (SHA3) and squeeze (extra length argument, only for Shake).

@gpshead the upstream PR will land soon, let me know if there's anything I can do to help make sure this gets into 3.12!

I've landed the upstream PR, meaning that this is now pulling from HACL*'s main branch. I see from https://peps.python.org/pep-0693/ that the first 3.12 beta is in just a few days. I'll keep an eye on this PR in case any change or fixup is needed. Thanks!

🤖 New build scheduled with the buildbot fleet by @gpshead for commit b15b524 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

At first glance I believe this looks good, I'll get it in. I'm running it through the buildbots to see if anything unusual build-wise falls out to be addressed. (expect some flaky infrastructure or pre-existing issue noise among those results as well)

Thanks, so far only the refleaks test seems relevant -- it says hashlib leaked. Is there any way to get a trace of the allocations that weren't freed? Nothing jumps out at me looking at the code right now, and I don't believe anything was done differently compared to the other hashes (for which there were no leaks issues).

EDIT: used OSX's leaks tool in conjunction with MallocStackLogging, but I'm getting allocation traces from within the interpreter, which isn't super useful. I'll keep trying to figure this out, but of course any pointers are appreciated. Thanks.

EDIT2: tried tracing the calls to malloc performed within HACL, but the addresses don't seem to match what leaks tells me has leaked... maybe the leaked objects are Python objects?

I merged master in and I think I found the missing PyBuffer_Release. Can you run buildbot again?

🤖 New build scheduled with the buildbot fleet by @gpshead for commit 43f4ba7 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

This is the only error that shows up (WASI build):

ERROR: test_with_segments (test.test_pathlib.PathSubclassTest.test_with_segments)
ERROR: test_with_segments (test.test_pathlib.PathTest.test_with_segments)
ERROR: test_with_segments (test.test_pathlib.PosixPathTest.test_with_segments)

This doesn't seem to be related, does it?

@AA-Turner I pushed a commit to enforce better ordering.

I'm not sure how the list is ordered, though: from _abc.c to _weakref.c are files starting with an underscore, correctly sorted. Then, there is a mix of underscore-prefixed files and non-underscore-prefixed files, sorted alphabetically, after removing the initial underscore: arraymodule.c, ..., _operator.c, posixmodule.c, and so on.

I did my best to try to make the list look sensible. I hope this addresses your comments, let me know if it doesn't. Thanks!

@zooba would likely be the best to comment on what should be the case, though your changes seem alright to me -- thank you.

A

Great, thanks. I'm also happy to do a followup PR to fix things up after this one lands, if need be.